AI 中文总结
该研究针对随机交通网络的车辆路径问题，提出基于量子幅度估计的行程时间估计方法，在IBM Qiskit上实现多种变体，实验表明其可实现优于蒙特卡洛的效率，为智能交通系统提供参数选择指导。
AI 中文摘要
智能交通系统（ITS）的核心任务是求解随机交通网络（STNs）中的车辆路径问题（VRP），该任务面临随机路径行程时间及由此产生的VRP目标函数的估计挑战。这些挑战通常通过计算成本高昂的基于采样的技术（如蒙特卡洛模拟）解决，其性能取决于样本量、采样策略和潜在的行程时间分布。为解决这些问题，本研究提出并验证了一种量子计算技术——用于STNs中路径级行程时间估计的量子幅度估计（QAE）。该框架不依赖采样或行程时间分布的先验假设，而是将所有可行的行程时间实现编码为量子叠加态，理论上可实现比蒙特卡洛模拟快一倍的加速。在IBM的Qiskit框架中实现了四种QAE变体，即经典幅度估计（CAE）、迭代幅度估计（IAE）、最大似然幅度估计（MLAE）和更快幅度估计（FAE），同时采用四种旋转角缩放策略以处理不同的离散行程时间分布。在小规模STN上的实验表明，缩放方法和旋转角范围的选择会显著影响估计精度，而四种QAE变体在所有测试条件下产生的估计值具有可比性，其中IAE表现出最稳定的整体性能。这些结果为ITS应用中未来的混合量子-经典优化框架提供了参数选择的实用指导。
英文摘要
Solving the Vehicle Routing Problem (VRP) in Stochastic Transportation Networks (STNs), a core task in Intelligent Transportation Systems (ITS), introduces estimation challenges for stochastic path travel times and the resulting VRP objective function. These challenges have typically been addressed through computationally expensive sampling-based techniques such as Monte Carlo simulation, whose performance depends on sample size, the sampling strategy, and the underlying travel time distributions. To address these issues, this study proposes and validates a quantum computing technique, Quantum Amplitude Estimation (QAE) for path-level travel time estimation in STNs. Without relying on sampling or prior assumptions of the travel time distribution, the proposed framework encodes all feasible travel time realizations into a quantum superposition, enabling a theoretical quadratic speed-up over Monte Carlo simulation. Four QAE variants are implemented in IBM's Qiskit framework, namely Canonical AE (CAE), Iterative AE (IAE), Maximum Likelihood AE (MLAE), and Faster AE (FAE), together with four rotation-angle scaling strategies for handling different discrete travel time distributions. Experiments on a small-scale STN show that the choice of scaling method and rotation-angle range significantly affects estimation accuracy, while the four QAE variants produce comparable estimates across all tested conditions, with IAE exhibiting the most stable overall performance. The results provide practical guidance on parameter selection for future hybrid quantum-classical optimization frameworks in ITS applications.
